parquet: decoder-level dictionary filter pushdown for Parquet reader#9464
Open
lyang24 wants to merge 1 commit intoapache:mainfrom
Open
parquet: decoder-level dictionary filter pushdown for Parquet reader#9464lyang24 wants to merge 1 commit intoapache:mainfrom
lyang24 wants to merge 1 commit intoapache:mainfrom
Conversation
lyang24
changed the title
lyang24
force-pushed
the
dict_decoder_predict_pushdown
branch
4 times, most recently
from
a4545dd to
644295d
Compare
Push predicate evaluation into the decoder for dictionary-encoded BYTE_ARRAY columns. Instead of decoding all rows into StringViewArray and then filtering, this evaluates the predicate once on the small dictionary (~N unique values), then maps integer keys to booleans via a simple lookup — no intermediate arrays are created for data rows. Two-phase approach: - Phase 1: decode dictionary page, evaluate predicate, produce matching_keys: Vec<bool> per row group - Phase 2: DictFilterDecoder reads RLE-encoded integer keys and maps each key to matching_keys[key], producing a BooleanArray that feeds directly into RowSelection Adds ArrowPredicate::use_dictionary_encoding() opt-in flag and falls back to the normal path for multi-column predicates, non-BYTE_ARRAY types, nested columns, or PLAIN-encoded pages.
lyang24
force-pushed
the
dict_decoder_predict_pushdown
branch
from
644295d to
53bd848
Compare
Contributor
Author
|
Hi @Dandandan, can you please help me summon the gh bots for clickbenches. |
Contributor
|
run benchmark arrow_reader_clickbench |
Contributor
Sorry, missed your message. Started now! |
|
🤖 |
|
🤖: Benchmark completed Details
|
Contributor
Author
|
interetsing i have the opposite results on q13 locally |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
Rationale for this change
For dictionary-encoded columns, we previously converted the dictionary data into a StringViewArray and then applied the filter. This approach materialized every row in the column before filtering, which introduced overhead.
What changes are included in this PR?
We introduced DictFilterDecoder for columns that are 100% dictionary encoded.
The decoder first processes the dictionary keys, then applies the filter.
After that, it expands null values.
Finally, it filters the dictionary values and materializes only the rows that pass the filter.
Are these changes tested?
existing test should pass added additional tests
Are there any user-facing changes?
yes added an new api with_dict_pushdown to opt in to enable the predicate pushdown on predicates